Barney Krishnan is a Data Executive at UniCredit with expertise in financial services, digital banking, AI and data modernization platforms.
getty
True software productivity requires retrofitting legacy software development life cycle (SDLC) platforms, redefining engineering metrics and harnessing the Jevons paradox.
In the early 1800s, as steam power swept through the British cotton industry, factory owners faced a deeply labor-intensive spinning market. Thousands of manual workers operated spinning wheels to transform raw cotton into thread.
When early spinning machinery emerged, industrialists attempted a one-to-one swap: replacing 100 workers with 100 machines. Yet, initial productivity gains were surprisingly marginal. Machinery worked tirelessly, but placing mechanization into a workflow designed for manual human labor created severe bottlenecks.
Real exponential productivity materialized only when mill owners rewired the physical assembly line. By introducing long, horizontal spinning frames that ran continuously across factory floors, they transformed fragmented workstations into an integrated system. It was not the machine alone that created the revolution—it was the systemic redesign.
The Trap Of Point Automation In Modern Software
Today, technology leaders stand at an identical inflection point with generative AI (GenAI). In the SDLC, many organizations repeat the 19th-century mistake: attempting a one-to-one replacement of isolated human tasks.
Companies often deploy AI coding assistants for developers or test generators for QA, expecting immediate transformation. However, I’ve found that replacing discrete tasks yields marginal overall gains. If an AI tool generates code five times faster, but product specs take weeks to clarify or QA manually reconciles broken architectural contexts, overall velocity barely shifts.
Replacing individual steps inside a legacy process is a point solution. Leaders must ask a different question: How do we rewire the entire software development lifecycle for native AI orchestration?
The Architecture Of AI-Driven Development
Rewiring the SDLC requires connecting every stage—from requirements and functional specs to code synthesis, automated testing and deployment. Bridging this lifecycle effectively demands two core components:
1. An End-to-End Execution Framework: You need an overarching framework that orchestrates tasks seamlessly, ensuring outputs from requirement engines flow directly into technical specs, code synthesis pipelines and deployment protocols.
2. A Universal Context And Knowledge Layer: This is the persistent connective tissue that captures, retains and distributes institutional memory across every phase—understanding infrastructure constraints and guidelines, past pull requests and legacy decisions to ensure AI agents make informed actions.
Design Is The New Code
As automated systems absorb tactical tasks like writing boilerplate code, drafting test cases and creating functional documentation, human talent generally shifts upstream. This evolution elevates a powerful imperative: Design is the new code. When routine execution is automated, human capital often shifts from line-by-line coding to high-value architectural design, domain modeling and strategic verification.
Historically, based on what I’ve seen in the industry, software engineers have spent a large portion of their time translating logic into syntax. In an AI-native ecosystem, engineers can pivot upstream toward high-order conceptualization—focusing on architecture/design patterns, documentation designed as structured, machine-readable Markdown files, user experience, state management and database modeling.
Humans can transition from manual authors to strategic designers. The engineer’s primary responsibility can then become setting boundary constraints, establishing guardrails and auditing AI-generated outcomes. Expertise can then be measured not by coding speed, but by system resilience and domain elegance.
Retrofitting Old Infrastructure: Preparing The SDLC For AI
When industrial steam power emerged, factory owners discovered they could not simply drop heavy steam engines into old timber-framed mill buildings without structural risk. They had to spend capital reinforcing foundations and re-engineering power shafts to support automated energy.
Technology organizations face an identical infrastructure challenge today. Dropping AI coding tools onto legacy, fragmented SDLC platforms often yields fragile results. To capture true returns, leaders must retrofit underlying infrastructure for native AI orchestration—rebuilding toolchains, reengineering applications, unifying API gateways/middleware and standardizing governance across the delivery pipeline.
Jevons Paradox: Why Automation Explodes Total Demand
A common misconception is that increasing automation decreases total labor output. History proves the exact opposite through the Jevons paradox. During the Industrial Revolution, as spinning machines drastically reduced the labor required per yard of thread, yarn costs plummeted. Lower costs drove an explosion in worldwide demand for textiles.
The textile industry’s total output surged, expanding the overall economic footprint. The same principle applies to software development today. Building on this century-old principle, generative AI will not decrease engineering work; it will expand it exponentially.
By reducing unit creation costs, I believe organizations will build previously cost-prohibitive custom tools, hyper-personalized applications and niche integrations. The volume of digital work will grow, requiring engineers to guide far larger, highly sophisticated software ecosystems.
Beyond Lines Of Code: Redefining Engineering Productivity
For decades, technology leaders evaluated productivity using metrics such as lines of code (LOC), commit frequency and story points. In an AI-native world, these metrics will become obsolete. AI can generate thousands of lines of code in seconds—making code volume a liability rather than an asset.
Engineering productivity must be re-anchored around the following outcomes:
• Architectural Integrity And Resilience: Organizations must measure how effectively systems maintain clean microservices, data modeling and scalar stability.
• Regulatory Compliance And Governance: Organizations must also make sure their AI-generated logic reliably aligns with legal, privacy and security standards.
• User Experience And Value Delivery: Organizations must determine whether software seamlessly satisfies end-user intent and business domain expectations.
User experience and system health—not raw code volume—will become the true indicators of productivity.
The Modern Factory Manager
Walk into a modern textile plant today, and you will not find thousands of laborers manually spinning individual threads. Instead, a small team of plant managers oversees an automated, high-precision operation, including temperature control, with a focus on system health, quality assurance and throughput optimization.
Software engineering is undergoing the exact same maturation. Organizations that view generative AI merely as a tool to automate isolated tasks will miss its true value. The leaders of the AI era will be those who courageously rewire their operational lifecycle—transforming software creation into a unified, AI-driven engine governed by human architectural brilliance.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?